Christopher Gillberg. Goteborg University, Sweden. This paper reports on the effect of using an interactive and child-initiated microcomputer program (Alpha) ...
Journal of Autism and Developmental Disorders, Vol. 25, No. 5, 1995
Increasing Reading and Communication Skills in Children with Autism Through an Interactive Multimedia Computer Program^ Mikael Heimann^ Goteborg University, Sweden
Keith E. Nelson^ Pennsylvania State University
Tomas Tjus Goteborg University, Sweden
Christopher Gillberg Goteborg University, Sweden
This paper reports on the effect of using an interactive and child-initiated microcomputer program (Alpha) when teaching three groups of children fN = 30) reading and communications skills: (a) 11 children with autism fM chronological age, CA = 9:4 years), (b) 9 children with mixed handicaps (M CA = 13:1), and (c) 10 normal preschool children (M CA = 6:4 years). Their mental age varied from 5:8 years to 6:9 years and all children received computer instruction supplementary to their regular reading and writing 'This research was supported by a grant from the Swedish Council of Social Research, Stockholm (Grant 90A)090) and two supplementary grants from the First of May Flower Annual Campaign for Children's Health, Goteborg, and from the Bracke OstergSrd Foundation, Goteborg, to Mikael Heimann. Portions of this paper have been presented at the second European Conference on the Advancement of Rehabilitation Technology, Stockholm, Sweden, May 1993. We gratefully acknowledge the assistance of Mats Lundalv, Margareta Kamevik, Eva Lundalv, Lisbeth B. Lindahl, and Jan Olsson in conducting this study. Special thanks are also due to all the participating children, their families, and their school teachers. ^Address all correspondence to Mikael Heimann, Department of Psychology, Haraldsgatan 1, S413 14, Goteborg, Sweden, or Keith E. Nelson, Department of Psychology, 414 Moore Building, University Park, Pennsylvania 16802. 459 0162-3257fl5/1000-(M5yS07.50/0 O 1995 Plenum Publishing Corporation
460
Heimann, Nelson, Tjus, and Gillberg
activities. Tests of reading and phonological development were carried out at the onset of the training (Start), at the end (Post I), and at a follow-up evaluation (Post 2). In addition, video observations of the childrens' verbal and nonverbal communication were added at Start and Post I. The children with autism increased both their word reading and their phonological awareness through the use of the Alpha program. Clearly significant gains were observed during the intervention, but none during the follow-up period. A similar but weaker pattern is observed for the children with mixed handicaps. In contrast, the normal preschool children increased their scores regardless of the program. Analyses of the children's classroom behavior indicate that the intervention succeeded in stimulating verbal expressions among the children with autism and mixed handicap. A significant increase in enjoyment was also noted for the children with autism. It is concluded that the intervention with a motivating multimedia program might stimulate reading and communication in children with various developmental disabilities, but that such interventions must be individually based and include both detailed planning and monitoring from teachers, and parents, as well as from clinicians in charge.
One of the earliest attempts to use a computer to stimulate language development in children with autism was described by Colby (1973). The program he used allowed the child to press a letter on the computer and simultaneously hear the computer say the letter. In another game, the child pressed a letter (e.g., "H") and then saw a horse moving across the screen together with sound from the horse's hoofs. The aim of this early multimedia attempt was to mimic normal spontaneous language acquisition and to encourage free exploration of the computer material. Colby reported on the results from 17 mute children with autism and claimed that 13 of these showed positive gains. That is, they started to use some voluntary speech and often also displayed enjoyment and motivational gains. However, no details were presented as to how many sessions were used, how long the intervention continued, and how the children were diagnosed. About a decade after Colby's report, Panyan (1984) pubhshed a review on the use of computers with children with autism. He noted that the computer technology offers greater possibilities for enhancing both interaction and attention, but that few systematic studies had been reported thus far. According to Panyan, computers coutd be used to address several areas relevant for people with autism, as for example: stimulus overselectivity, motivational support, and for improving interaction.
Facilitating Language Skills
461
Today, more than 10 years after Panyan's review, we still lack good systematic observations on the effect of computers for stimulating learning and/or communication among children with autism. One of the few findings reported in the literature has been provided by Jordan and Powell (1990a, 1990b) who reported positive effects of an intervention program that focuses on enhancing different cognitive skills (e.g., problem-solving strategies) among children with autism: The British children that used their program displayed positive gains as measured by cognitive tests. However, they also stated that children with autism need to be encouraged (not directed) in order to keep attention, and that the tasks must be "seen as solvable by the child" (Jordan & Powell, 1990a, p. 22). Positive effects are also reported by Bernared-Optiz, Ross, and Tuttas (1990) who used computer-aided instructions (CAI) with autistic subjects to facilitate learning in school settings both in Germany and Singapore. One of their studies compared the effectiveness of computer-aided learning with more traditional learning relying on personal instruction. Of 18 observed children and adults with autism (ages 5-31), 6 were rated higher when learning through CAI while only 1 was rated highest when personal instruction was used. Furthermore, they aiso described how CAI could help a 17-year-oid boy to use less echolalia and how a 16-year-old boy with poor writing skills managed to increase his skills substantially through a computer intervention. A third attempt to use CAI with children with autism has been presented by Coldwell (1991a, 1991b), an Australian researcher who has studied computers as a means of communication for mute autistic children. The hypothesis behind this attempt is that mute children with autism develop their own graphic symbols that are understandable to themselves and often also other children with autism, but not to nonautistic subjects. In one of his studies, Coldwell reported on eight children with autism in the age range 3-12 years (2 giris, 6 boys) who worked with the computer together with their parent with the aim to use graphics and to create symbols of various kinds. According to Coldwell, several subjects displayed concentration and responses that went far beyond what was expected based on former performance and he claims—or hopes—that, in the future, people with autism might be helped through computer networks and data banks to communicate with each other through a graphic communication system Sf>ecially developed by and for them. Although Coldwell's expectations might be judged as both too speculative and too provocative, the attempt to use CAI to support language and communication growth for disabled children is still a hopeful avenue for present and future educational attempts. We know that it is absolutely essential for a child's future functioning that the child has been given adequate possibilities to develop his or her language skills as far as is possible. This holds true not only for children with developmental disabilities like autism
462
Heimann, Nelson, Tjus, and Gillberg
(Howlin & Rurter, 1987; Schopler & Mesibov, 1985), but also for children with special needs (e.g., Baldrey, 1991; Douglas, 1991; Hasselbring, 1984; Semmel, Cosden, Semmeh, & Keleman, 1984) and normal children (e.g.. Nelson, 1977). It seems therefore extremely important to increase our tools for helping children develop language skills. To date, several studies report on the positive use of computers for children with various developmental disabilities (e.g., Green & Clark, 1991; Light, 1988; Nelson, Prinz, Prinz, & Dalke, 1991; Romski & Sevcik, 1989; Underwood & Underwood, 1990), but solid data on the effectiveness of computers for children with autism are still lacking. Very few studies have presented observations from well-designed experiments or quasi-experiments and several of the reports also failed to report enough details as to how the actual training was carried out. Finally, there is also a need to address the issue of individual strategies when learning language. Normal children use different strategies when learning their native language (Bates, 1979; Nelson, 1991, Nelson, Baker, Denninger, Bonvillian, & Kaplan, 1985) and this is probably also the case for children with autism as well as for children with other types of developmental disabilities (Howlin, 1989; Iacono, 1992; McTear & Conti-Ramsden, 1992). Furthermore, Romanczyk, Ekdahl, and Lockshin (1992) reported that they often have observed that even those children with autism that have explicitly expressed their preference for the computer tend to actually perform better with the teacher. Thus, we should not expect that CAI and related interventions will solve all the problems for children with autism. Rather, it is our view that computer-aided interventions might be of some help to some children with this diagnosis. This paper presents observations and results from a quasi-experimental field study aimed at investigating the effects of CAI that included a highly motivating and interactive multimedia environment when teaching children with autism reading and writing skills. A Swedish version of Alpha (Alpha Interactive Language Series/Gator Super Sentences; Nelson & Prinz, 1991) was used for teaching reading and communication skills to (a) children with autism, (b) children with various degrees of cerebral palsy and/or mental retardation, and (c) a group of normal preschool children. The program uses on-screen animations as well as videodisc material that gives the child an immediate feedback. Each noun or verb is immediately animated during sentence creation, and after completion the whole sentence is shown in text and as an animation. In addition, the teacher promotes a warm and supporting atmosphere and uses recasts, questions, and elaborations that tie in with the child's verbal activity and thus promotes a learning environment that maximizes the likelihood for the child to learn new language structures (see Nelson, 1991).
Facilitating Language Skills
463
Hypotheses Previous research with the Alpha program reports positive effects on the reading and language developmetit for children with severe hearing impairments (Prinz & Nelson, 1985; Prinz, Pemberton, & Nelson, 1985), children with multihandicaps, normal preliterate children (Nelson, Loncke, & Camarata, 1993), and also for a subgroup within the curretit project (Heimann, Nelson, et al., 1993a, 1993b). Despite the obvious differences between these groups, under conditions of CAI support for appropriately challenging lessons they all show a pattern of significant gains in reading skills. Thus, we developed specific hypotheses regarding (a) the children's reading development, (b) the children's verbal expressiveness while interacting with the teacher, and (c) the children's motivation for communication. Furthermore, we also predicted positive changes within two areas of communicative development that could be facilitated by high attention to the computer material in conjunction with the teacher's active dialogue with the child: Phonological awareness and overall verbal level as measured by a sentence imitation test. Positive changes were expected within all these areas as an effect of the CAI as implemented with teachers as active partners.
METHOD Subjects A total of 30 children divided into three groups participated in the study. Children with Autism. Group A children were 9 boys and 2 girls diagnosed according to the DSM-III-R (American Psychiatric Association, 1987) as suffering from autistic disorders, and with a chronological age between 6:9 and 13:8 years (M age = 9:4; see Table I). Nine of the children were diagnosed at the Child Neuropsychiatric Clinic, Goteborg (Head: C.G.) and two at a regional hospital (Lidkoping, Sweden). Mental age varied between 3:0 and 9:5 years (Coloured Progressive Matrices, a test of nonverbal intelligence; Raven, Court, & Raven, 1984) and receptive language age between 2:9 and 7:0 years (estimated by an adaptation into Swedish of the Norwegian version of the Reynell Developmental Language Scales; Reynell, 1977; Hagtvet & Lilliest0len, 1984). All children attended school clinics specialized for teaching children with autism and all children were judged by their teacher to have displayed some basic communicative motivation.
464
Heimann, Nelson, Tjus, and Gillberg Table I. Mental Age (MA), Chronological Age (CA) and Language Age (LA) for the Participating Children" MA
CA
LA
Groups
n
M
SD
M
SD
M
SD
Combined (A + MH) Autism (A) Mixed handicap (MH) Normal preschool
20 11 0 10
6:3 6:9 5:8 6:3
1:8 2:1 0:5 1:2
11:4 9:4 13:1 6:4
4:6 2:4 5:6 0:8
4:7 4:9 4:1 6:10
1:6 1:10 1:3 0:3
" Age given in years:months.
Children with Mixed Handicaps. Group MH comprised 9 children (4 boys) with a mean chronological age of 13:1 years {SD = 5:6). Their estimated mental age ranged from 5:0 to 6:6 years (see Table I) and the observed receptive language age ranged from 2:8-7:0. All were judged to have an IQ score of 70 or less. Seven of the children had at least one motor or sensory impairment and 2 of the children (2 boys) had received the diagnosis Down syndrome. Normal Preschool Children. Group NP consisted of 10 normal preschool children (2 boys) enrolled in a normal day-care institution. Their mean chronological age was 6:4 years and their mean mental age 6:3 years. Their language development ranged from a language age of 6:3 to 7:0 years.
Procedure All children used a Swedish version of Alpha and received a number of 19.3 training sessions {SD = 11.9) during 3 to 4 months {M = 13.6 weeks, SD = 8.3). More specifically. Group A received on average 25.6 sessions {SD = 7.5) over 16.9 weeks {SD = 5.7); group MH, 21.8 sessions {SD = 12.7) over 17.7 weeks (SD = 10.1); and Group NP, 7.8 sessions {SD = 7.3) over 6.3 weeks {SD = 2.5). The training was carried out weekly {M = 1.5 sessions per week; SD = 0.9) and each session lasted between 21.1 {SD = 5.0; Group MH) and 32.0 minutes {SD = 12.6; Group A). All children were given a familiarization period with the Alpha program prior to the actual training. The purpose of this period was (a) to see if the children were as motivated and as interested as had been indicated by their teacher and/or patents, (b) to allow all children to learn the basic set-up and functions of the program, and (c) to find out the right level to start from. The number of familiarization sessions were 5.9
Facilitating Language Skills
4^
{SD = 3.3) for Group A, 13.0 (SD = 7.9) for Group MH, and 3.2 (SD = 2.4) for Group NP. The children were tested at three occasions and observed (videorcorded) twice. The first test was carried out after the end of the familiarization period {Start test), the second test (Post 1) during the last week of training, and the foliow-up assessment (Post 2) approximately one semester (M = 26.2 weeks, SD = 14.8) after Post 1. Two of the children in the normal preschool group did enter school during their follow-up period.
Program Description
The Alpha program (Nelson & Prinz, 1991) is constructed to facilitate language learning through multichannel feedback (voice, animation, video, and sign language). For this study, sign language was not used and Alpha was translated to Swedish using the words and animations already incorporated in the United States version. Thus, no attempt was made to change the program towards how familiar or difficult the words were to a Swedish child learning to read her or his native language. The Swedish Alpha consists of 112 lessons, all aimed towards developing a basic reading and writing vocabulary, and the ability to create simple sentences using this vocabulary. The program makes it possible to select one of four main working modes: Individual Words (IW), Creating Sentences (CS), Testing Words (TW), and Testing Sentences (TS). Initially all children started with the IW mode in order to learn the vocabulary (= nouns) of a particular lesson. When the child mastered the lesson, as indicated by the test score (TW), the child moved on to the CS mode. Within this mode a child could create sentences by combining earlier learned nouns with new verbs. That is, selection of a simple noun-verb-noun sequence like "The bear"-"jumps over"-"the horse" creates an animation showing the action the child has described. Feedback is also provided during sentence construction. Each noun is illustrated by an appropriate animation while each verb is illustrated by two different animations illustrating the same action. Those children using a system version that included a videodisc player (see Equipment, below) saw the nouns and the verbs illustrated by short video examples instead. After having explored a lesson in the CS mode twice (or earlier if it was obvious that the child had mastered the level) the teacher switched to the TS mode for that lesson. In this mode, the program produced an animated depiction of a noun-verb-noun sequence randomly generated from the vocabulary used in the lesson. After having watched the animation, the child had to select the nouns and the verb to construct the sequence that described what the ehild had just seen. At any given level, a child was judged
466
Heimann, Nelson, Tjus, and Gillberg
to have reached mastery if reaching a score of or above 80% correct in the TS mode. Measures
Several tests were used to assess each child's language and communication skills before and after the training period: Reading. Three different tests were used in combination order to measure progress in reading: Flashcard A or B (sentences), Flashcard C (words), and Umesol (letter identification and word reading). Flashcards A, B, and C were all developed within the project while Umesol (Taube, Torneus, & Lundberg, 1984) is a well-known test in Sweden. The reading score used in the analysis is based on a combined score from all three tests. Sentence Imitation. This test, based on previous research with the Alpha program (Prinz et al. (1985), measures the child's ability to imitate in his or her best communicative mode (i.e., spoken language, Swedish Sign Language, or BLISS symbols). It includes items of increasing length and grammatical complexity. Phonological Awareness. The children's phonological awareness (sound synthesis) was assessed using a Swedish instrument (Torneus, Taube, & Lundberg, 1984). Communication. Video recordings of each child's communicative behavior during one initial and one final lesson were analyzed using five different categories: Complies, Off Task, Seeks Help, Verbal Expressions, and Enjoyment (a complete description of the coding criteria and procedure can be obtained from the first author). A total of 9 minutes, divided into three periods (first, middle, and last 3 minutes of the session), were coded for each lesson and the occurrence of each category during every 10-second interval was noted by the coder. One of the authors (T.T.) coded all observations (a total of 38 lessons for groups A and MH) and rehability was checked by having a graduate student code 10% of the material (randomly selected). The obtained overall reliability coefficient was .86 for Pearson's r, and .81 for kappa (Cohen, 1960). Level of Autism. The level of autistic features/behaviors among the children diagnosed as autistic (Group A) and the children with mixed handicaps (Group MH) was estimated by completing the Childhood Autism Rating Scale (CARS; Schopler, Reichler, & Renner, 1988). All children in both groups were rated from videotape and through interviews with the child's teacher by a psychologist who had not been part of the original study and who was also uniformed as to what group the children belonged. It was found that
Facilitating Langua^ Skills
4£7
the average scores for Group A (M = 41.9; 5D = 117) and Group M H ( M = 26.1; SD = 4.4) differed significantly, t(lS) = 3.55, p < .01. Equipment Each child worked with either an Apple He or Apple IIGS microcomputer system with a minimum of 12 8 K RAM and a printer. In addition, all children in the comparison group and seven of the children with autism had access to a Sony videodisc player (LDP1-3600D). Subject Loss Children with Autism. Group A origjnalfy consisted of 12 children, but 1 child neier am^ieted the training paiod due to froblems in arranging an acceptable educational environment. In addition, 2 more children were dropped during the familiarization period; they were not attracted by the computer and they also di^layed a wide variety of severe stereotypical and uncontrolled destructive behavior. Children with Mixed Handicaps. Group MH: One girl left the study after only a couple of weeks of training. She was motivated by the program but became so frustrated by her slow progress that her teacher and her parents decided not to continue. Normal Preschool Children. Group NP: One additional child started the training, but never had the chance to continue for an acceptable length of time due to her family leaving for an early summer vacation. Data Loss
,
Not all children in each group were successfully tested on all tests. For Group A the actual number of children included varied from 7 (sentence imitation) to 11 (coding of communication) while for Group MH the variation was between 5 (sentence imitation) and 9 (coding of communication). For Group NP, several children were lost after completion of the training, thus making it impossible to carry out any follow-up test. As an example, successful video observations could be carried out for only 3 children in Group NP when ending their training. This large and unejq^ected loss of data was due to administrative and political reasons: The city council of Goteborg drastically implemented strong cutbacks severely affecting the participating day-care center. Statistical Analysis The analysis focuses on changes over time in observed means within each group/subgroup. Since both parametric (paired / test) and nonparametric methods (Wilcoxon) reveal similar results only the ( statistics are reported in the present paper. Furthermore, a one-tailed significance level
468
Heimann, Nelson, Tjus, and Gillberg Table IL Number of Alpha Lessons Covered from Start to Posttest 1 (Max = 112) Alpha lesson number Start : test
Combined (A + MH) Autism (A) Mixed handicap (MH) Normal preschool
Posttest 1
M
SO
M
6.1 4.9 7.3 2.6
6.9 6.7 7.2 1.9
21.1 22.2 19.9 21.5
Comparison
SD
df
(
12.1 15.6 8.2 11.2
17 8 8 7
5.4" 3.5" 5.3"
"p < .01. '•o < .0001.
is accepted throughout since each of the results refers to one of our specific hypotheses. No systematic between-groups comparison has been carried out due to small Ns and to fewer lessons given to the normal preschool children as compared with the children with autism and mixed handicaps.
RESULTS I
The Alpha Program
All three groups of children made considerable and significant progress within the Alpha program from the onset (Start) to the end of the training (Post 1). As presented in Table n, the children with autism progressed from Alpha Lesson 5 (M = 4.9, SD 6.7) to Lesson 22 (M = 22.2, SD = 15.6) while the children with mixed handicaps on average went from Lesson 7 (M = 7.3, SD = 7.2) to Lesson 19 (M = 19.9, SO = 8.2). The normal preschool children showed a similar pattern despite the fact that they received a shorter intervention period: They progressed fiom Lesstxi 3 (M = 16, SD = 1.9) to Lesson 72 {M = 215, SD = 112). Furthermore, the total number of lessons that the children received, that is the sum of both the initial familiarization period and the subsequent intervention period, varied narrowty ftom a mean of 20.0 {SD = 8.2) for Group MH to 21.4 (SD = 11.7) for Group A, and a mean of 26.1 (SD = 11.4) Alpha lessonsforGroup NP. A post-hoc comparison between the three groups using Studenfs f test showed a nonsignificant result (p > .10). Thtis, the three groups seem to have received very similar amounts of training as indicated by the number of lessons covered within the Alpha program. The children's performance on Alpha's built-in test modes (Testing Words and Sentence Testing) indicate that all three groups did learn both words and sentences through the program. The actual performance on the final sentence test show that the children with autism received a score of 92% correct (SD = 8.4), the children with mixed handicaps scored 80.9%
Facilitating Language Skills
469
{SD = 11.1), and the nonnal preschool children 97.9% correct {SD = 6.0). Thus, the children in all three groups had learned to master new language material during their training with Alpha. Reading A clearly significant change, f(15) = 2.88, p < .01, is noted for the combined results for the children with autism (Group A) and the children with mixed handicaps (Groups MH) during the training period with Alpha (from Start to Post 1; Table III). Stated more specifically, the children with autism increased their mean scores from .03 to .14, /(8) = 2.85, p < .05; the children with mixed handicaps from .18 to .23, t{l) = 1.79, p < .06; and the normal preschool children displayed a gain of .16, r(9) = 2.11, p < .05. The children in Group A also displayed a significant change from Start to the follow-up evaluation (Post 2), t{8) = 2.85, p < .01, but no significant change is noted for the follow-up period per se. The children in Group MH displayed no significant changes when their result at follow-up (Post 2) is analyzed although a clear change in the observed means can be noted. In contrast to the other two groups, a significant change is noted for the normal preschool children during the follow-up period between Post 1 and Post 2 without Alpha teaching. This indicates that the Alpha program had a specific effect on the reading development for the children in Group A and Group MH while the children in Group NP showed a less specific response to the intervention. However, it ought to be noted that the changes in observed means, specifically for Group MH, also indicate strong individual variation among the observed results for the children.
Phonological Awareness
Overall, all children increased their means during the training period (Start-Post 1) which is evident by the combined result for Group A and Group MH displayed in Table IV. Significant gains are observed both during the training period proper, r(13) = 2.7, p < .02, and from Start until Post 2 at the follow-up evaluation, t{13) = 1.99, p < .05. If each group is analyzed separately, we note significant gains for both Groups A and NP from Start to Post 1 (see Table IV), t{l) = 2.5 and /(8) = 2.48, respectively, p < .03), and for Group MH from Start to Post 2, /(5) = 2.29, p < .05. In contrast with the results for the children with autism and the
Heimann, Nelson, Tjus, and Gillberg
470
U
CQ
U
u
CQ
(^
-- m
r-.
< I
< s 1/3
r^
OS CO
X
1
O
o
i I
Facilitating Language Skills
471
Table V. Start and Posttest Results for Sentence Imitation Start
Post 1
/-Test
Groups
n
M
SD
M
SD
(P
